On 12 Dec 2012, at 05:20, Chris Keown wrote:
Hello,
I have a dataset where the EPI was required obliquely. In my analysis, I am looking at local connectivity, so I am trying to minimize resampling and spatial smoothing effects and would prefer not to use AFNI's 3dWarp -oblique2cardinal. Is it safe to plug in oblique data to FLIRT, hence doing the registration and deobliquing all in one stuep? Also, do I further need to run 3drefit -deoblique to eliminate the obliqueness from the header file, and if so, before or after FLIRT?
In the general FSL pipelines (FEAT, MELODIC, etc) there is no need to deoblique in preprocessing - registration (eg between functional and structural) is driven by the image data and the native-space analysis runs fine without deobliquing.
Lastly, any other ideas on how to minimize spatial smoothing would be greatly appreciated. FNIRT/FLIRT w/ sinc interpolation worked well for registration, and AFNI Fourier interpolation seemed to work best for minimizing smoothing during motion correction.
That all makes sense - there is also a -spline_final option in mcflirt which would help too - though you would have to call that from the command line to get it as it's not a FEAT/MELODIC GUI option.
Cheers.
Thanks in advance!
chris